Company
Wio Bank PJSC, the Middle East’s first digital financial platform, is reimagining banking for individuals and businesses. Headquartered in Abu Dhabi and backed by strategic investors including ADQ, Alpha Dhabi, e&, and First Abu Dhabi Bank (FAB), Wio combines cutting-edge technology with a human-centric approach to deliver seamless, personalized financial solutions.
Challenge
Wio Bank wanted to increase its profitability by improving the efficiency of its cloud infrastructure and reducing cloud costs while maintaining performance. The company initially used cost monitoring and recommendation solutions but soon found that they lacked the ability to automatically apply recommendations and cut cloud costs in real time.
Solution
Starting with a detailed review and audit of Cast AI’s cost monitoring and recommendation features, Wio Bank first enabled automation in its non-production environment to test the solution. Next, the team confidently transitioned to its production environment and became more proactive in optimizing the non-production environment using automation tools and insights.
Results
- Up to 70% cloud cost savings per year per cluster
- Achieving 70-100% memory utilization
- 90% of non-production workloads running on Spot VMs for greater savings
Node rebalancing
Rebalancing is a Cast feature that automatically replaces suboptimal nodes with new ones that are more cost-efficient and run the most up-to-date node configuration settings. Cast maintained the optimal Spot VM cost by running nightly Scheduled rebalance jobs.
Pod bin-packing
Pod bin-packing occurs thanks to Evictor, which continuously compacts pods into fewer nodes, creating empty nodes that can be removed to achieve optimal CPU and resource utilization.
Implemented together, Cast’s Evictor and Rebalancer dramatically increased resource efficiency.
Full Spot VM Automation
Thanks to Cast, Wio Bank can now run 90% of its non-production environments on Spot VMs, unlocking massive cost savings.
Workload autoscaling
Workload Autoscaler automatically scales workload requests up or down to ensure optimal performance and cost-effectiveness. This feature helped Wio Bank achieve even greater cost savings following Spot VM automation.
In the graph below, the dark blue line represents resources provisioned for a given workload, the light blue field shows resources requested and the green one resources that ended up being utilized – all together showing the level of overprovisioning Cast helped to reduce.
The Workload Autoscaler had a significant impact on memory utilization, illustrated in the dramatic drop in requested GiBs:
By not having to spend time managing Kubernetes clusters, we can focus on other priorities and bring more capabilities into our environment, which is our main focus right now. The time saved thanks to Cast AI is being reinvested into other areas of our business.
Yannick Janssen, Deputy Chief Technology Officer at Wio Bank
Managing costs of a demanding cloud infrastructure
What goals must your cloud infrastructure meet to support your business operations as a financial services company?
In terms of our goals, we aim to support the business by delivering new features and capabilities to our customers. Our role is to maintain the underlying infrastructure, ensuring that it remains highly available and stable. One of our key objectives is to provide more stability so that our customers can enjoy uninterrupted access to our application, 24/7.
Additionally, we are actively involved in supporting Wio Bank PJSC as well as Wio Securities LLC. This is an important investment on our part, especially given that we are working with various virtual assets. Since trading occurs around the clock, the environment must be extremely stable to accommodate 24/7 activity.
Another critical aspect of our work is ensuring that our infrastructure remains cost-efficient. While the infrastructure is seen as a cost center – and indeed it is – controlling and reducing costs is essential to improving profitability.
What was your approach to cloud cost management? Have you used cost monitoring solutions?
We started early in our journey to manage costs, recognizing that cloud expenses can rise significantly based on usage and consumption. To address this, we initially used cost monitoring tools provided by the cloud providers. These tools come by default and offer various ways to help optimize cost management within the cloud environment.
We explored open-source technologies, many of which are already in use at Wio. These technologies were used to analyze and report on costs.
However, what we found lacking was a technology that would allow us to take direct action on cost management. While there are plenty of tools available for reporting and providing an overview of costs, we needed a solution that would enable us to act on those insights in real time.
What made you decide that cost monitoring is not enough to optimize cloud costs?
We received recommendations from various sources, particularly based on our usage of the environment. At a certain point, we realized that we needed to move beyond just the reporting phase.
Our focus shifted towards finding technologies that would not only help us report on costs but also actively reduce them. We implemented a function within Wio that was centered on action rather than just reporting.
The real goal was to find technologies that could trigger actions to directly control and reduce costs, ensuring that we had the tools to manage expenses more effectively.
Yannick Janssen, Deputy Chief Technology Officer at Wio Bank
Initially, we started with Cast to get insights into our environment – identifying areas where we could improve and assessing the potential value of those improvements.
After that, we shifted our focus to taking real action based on those insights. The key phase for us was figuring out how to apply the discoveries from Cast directly to our environment. It was crucial for us to move from simply identifying cost-saving opportunities to actively implementing them.
Turning to automated optimization
What was onboarding Cast like?
Early on, we worked closely with the Cast team to implement their technology within our cluster. It was a highly collaborative effort between our teams, and it resulted in a strong partnership. Together, we decided to move forward with automation, ensuring that we could take meaningful action based on the insights we gathered.
After conducting an initial review and audit of the environment, we chose to implement automation in our non-production environment first. We were cautious about applying any actions directly to the production environment without thorough testing. By running tests in the non-production environment, we ensured that everything was stable before moving forward.
Once we achieved success multiple times in the non-production environment, we confidently transitioned to the production environment. After that, we became more proactive and aggressive in managing our non-production environment, using insights and automation tools to optimize it further.
What have the results been so far, and what annual savings are you looking to get with Cast?
Looking across our environments, we have achieved up to 70% cost savings. For us, the cost savings have been a gradual process. Over time, we’ve steadily increased our savings as we’ve gained more experience and better optimized our usage with Cast AI.
Yannick Janssen, Deputy Chief Technology Officer at Wio Bank
At this point, we’ve been able to achieve approximately tens of thousands of dollars in monthly savings. This continuous growth reflects both our increasing usage and our deepening understanding of how to leverage Cast effectively.
What was your ROI timeline after implementing Cast?
For us, the results and savings came quite quickly, much faster than we expected. The savings were gradual, however, within the first three months of using Cast, we were already seeing the results we were aiming for. We believe that Cast delivers fast results, depending on how quickly you implement the solution and drive its usage.
Kubernetes automation in action
Which Cast features were the biggest game changers for you?
Node rebalancing and bin-packing
There were several phases in our cost-saving efforts. The first phase involved using Cast’s Rebalancer, along with Evictor. These worked together to identify underutilized nodes. The Evictor would move pods to different nodes, allowing us to terminate the underused nodes, which resulted in significant savings.
Leveraging Spot VMs
The second phase, where we experienced even more cost savings, was moving to Spot VMs. Currently, around 90% of our workloads in non-production clusters are running on Spot VMs.
Workload Autoscaler
After moving to Spot VMs, we discovered Cast’s Workload Autoscaler. Before that, we had a custom implementation of autoscaling, but it didn’t work as well or as flexibly as Cast’s alternative. As a result, we quickly decided to move to Workload Autoscaler and decommissioned our custom solution.
Workload Autoscaler has been a major contributor to our cost reductions, likely accounting for about half of the savings. This is because, by default, we had been overprovisioning resources like memory and CPU to avoid issues such as CPU throttling, which we had experienced in production.
However, with Workload Autoscaler, we have been able to stabilize resource allocation in the non-production environment. We plan to expand it to our production environment.
Oleksandr Zosyk, Senior DevOps Engineer at Wio Bank
How was the support?
I just want to emphasize again how amazing the support has been. It’s exceptional. When I talk with my friends about DevOps experiences or third-party providers, I always mention Cast and the incredible support you provide. It’s truly incomparable to what we’ve experienced with other third parties. I won’t name any, but your support stands out.
Honestly, sometimes we might ask questions that seem a bit silly, but your engineers are always willing to take the time to explain things to us. That level of dedication is amazing. What’s also impressive is how quickly we get responses – often within just a few minutes.
Oleksandr Zosyk, Senior DevOps Engineer at Wio Bank
Do you see Cast impacting your engineer workload as well?
The solution we found with Cast has helped us mitigate this issue of over-provisioning and paying more than necessary. Additionally, managing the cluster and replacing nodes – especially the automated replacement of nodes – was previously a painful process when done manually, even with the limited automation we had.
With Cast, we’ve achieved much more automation. It feels like having our DevOps engineer working alongside us in the background, helping us optimize processes that would otherwise take significant time from our team. Automating these tasks has been crucial for us. We want to do things more intelligently and efficiently, and this level of automation is exactly what we need. The team’s offer has been invaluable.
Yannick Janssen, Deputy Chief Technology Officer at Wio Bank
Automation with policies and controls
Did you have any doubts about delegating the task of cloud resource management to an automation engine?
We’ve had discussions with other companies – some very large ones – that were hesitant to implement Cast. They were particularly concerned about the potential impact of some of the actions on their non-production environments, which led them to stop or avoid moving forward. We experienced some of that fear ourselves at the beginning, but we worked closely with Cast, learning about the technology and how to make it safer.
For us, it’s been a very successful journey. We think other well-known technical teams who haven’t implemented Cast due to this fear are missing out on something valuable. We would like to take this opportunity to thank both the Cast team as well as our team at Wio Bank. Their collaboration has been essential in delivering the great results we’ve achieved.
What kind of controls have you used for Cast to ensure safe automation?
What we really appreciate about Cast is the range of functionalities and capabilities it offers, which allows us to make different decisions for production and non-production environments.
We can be more aggressive in managing resources in a non-production environment while being extra careful with production. This flexibility is one of the key aspects we like about the system.
Yannick Janssen, Deputy Chief Technology Officer at Wio Bank
For our non-production environment, we’re particularly happy with the “chaos monkey” approach we’ve implemented. It forces our developers to think critically about how their services behave when something goes wrong. They learn to make their services more fault-tolerant, which is incredibly valuable.
Even though it creates a bit of chaos, it’s actually very beneficial for the team. It challenges them and helps improve the overall reliability and resilience of the services they’re developing. It’s a great way to enhance the ability of the services and ensure they’re more robust in the long run.
Reinvesting the saved time and money
What is the broader impact of the cost and time savings Cast generated for Wio Bank?
The time we’ve saved has helped us move to the next level of what we’re doing. By not having to spend time managing Kubernetes clusters, we can focus on other priorities and bring more capabilities into our environment, which is our main focus right now. The time saved thanks to Cast is being reinvested into other areas of our business.
In the beginning, the time we gained from working with Cast also allowed us to invest in new features that was being delivered. We’ve always asked for new features, and we’ve noticed that the team has been very responsive, quickly building and adapting features based on our requirements and the needs of the customers.
This saved time has been reinvested in enhancing our operations and moving forward with the next steps. On one side, we’re working with Cast to improve and help everyone, and on the other, we’re exploring and investing in new technologies to embed into our environment.
What would you name as the three biggest benefits of Cast for Wio Bank?
For me, the main benefits have been cost savings, time freedom, and automation capabilities. Cost savings have been huge, and the ability to automate processes has freed up a lot of time for us to focus on other priorities. Automation has also played a key role in strengthening the availability of our environment. The automation provided by Cast is far superior to what we’ve experienced with the automation tools offered by Azure or other cloud providers.
Oleksandr Zosyk, Senior DevOps Engineer at Wio Bank
A partnership for the future
What has been your experience working with the Cast team, and what aspects of the partnership stand out to you the most?
For us, one of the most important aspects has been the support we’ve received from Cast team. The team is highly technical, which makes it a pleasure to work with people who truly understand the engineering side and aren’t just here to sell a product. It’s been a great journey and we hope it’s been just as positive for the team.
We would definitely recommend anyone to work with the Cast team because you’re doing fantastic work. You really listen to our needs, especially when it comes to new features and capabilities. We greatly appreciate this partnership.
